Lattice Boltzmann Model for Fast Level Set Algorithm Using the Multiple Kernel Fuzzy C-Means

Over the last decades, the development of high dimensional large-scale imaging devices increases the demand of fast and accurate image processing techniques. Due to its intrinsic advantages which allows to handle complex shapes and topological changes, the level set method (LSM) is an promising technique but is computational expensive. As a fast alternative approach for solving the level set equation, the highly parallelizable lattice Boltzmann method (LBM) has attracted much attention. Nevertheless, nearly all the level set image segmentation methods based on LBM employ the Bhatnagar-Gross-Krook (BGK) collision model. In this paper we firstly demonstrate that there is no necessary to use the BGK model in the level set image segmentation and the method is faster with almost the same result when considering zero collision. Experimental comparison results with four image segmentation methods based on level set using LBM-BGK confirm this theoretical analysis. We secondly propose a fast and efficient LBM-based level set method which incorporates intensity and texture information of the image. From the partition matrix of a multiple kernel fuzzy c-means (MKFCM), we design a multiple kernel fuzzy stop function (MKFSF) for the LBM solver based on the zero collision model. The method is fast, accurate and highly parallelizable. Experiments on natural and medical images demonstrate the superiority of the proposed method in term of speed and efficiency comparing with five image segmentation methods based on level set.